[Lucid][CVE-2013-0160][PATCH v2 0/7] TTY: do not update atime/mtime on read/write

Luis Henriques luis.henriques at canonical.com
Fri Jan 24 14:17:18 UTC 2014


And here's v2 after Andy's review.

This set includes 7 patches: 3 already present in v1, 2 identified by
Andy, another one (a clean cherry-pick) required to introduce
FMODE_NONOTIFY, and an UBUNTU patch to fix ubuntu/aufs as there was an
interface change in fsnotify_open().

Here's the list (upstream SHA1s):

* 2a12a9d "fsnotify: pass a file instead of an inode to open, read, and write"
  Just a pre-req to the next patch
* "UBUNTU: ubuntu: AUFS: fsnotify_open() now receives a file"
  Since the previous commit changes the fsnotify_open() interface,
  AUFS had to be changed
* ecf081d "vfs: introduce FMODE_NONOTIFY"
  Dropped a hunk in include/linux/fs.h: OPEN_FMODE isn't defined in
  Lucid and the commit that introduced it (5300990 "Sanitize f_flags
  helpers") didn't seem to be worth backporting.
  Also, the inotify_inode_queue_event() calls in
  include/linux/fsnotify.h (not present in the original commit) are
  now conditional.
* 12ed2e3 "fanotify: FMODE_NONOTIFY and __O_SYNC in sparc conflict"
  Clean cherry-pick
* b0de59b "TTY: do not update atime/mtime on read/write"
  Same as v1
* 37b7f3c "TTY: fix atime/mtime regression"
  Same as v1
* b0b8856 "tty: fix up atime/mtime mess, take three"
  Added the FMODE_NONOTIFY-related code in drivers/char/pty.c (note
  the file rename), that was missing in v1

In addition to the testing I had done already for v1 (running the PoCs
and the 'w' utility), I've also executed a few inotify tests to make
sure I wasn't breaking anything:

* Executed inotifywatch on /dev/ptmx device to verify that no events
  occur when accessing this device (they are being delivered using
  current Lucid kernel)
* LTP inotify and inotify_init syscall tests
* a little bit of stress testing using iwatch to recursively watch a
  directory while fs activity was going on


Eric Paris (2):
  fsnotify: pass a file instead of an inode to open, read, and write
  vfs: introduce FMODE_NONOTIFY

Jiri Slaby (2):
  TTY: do not update atime/mtime on read/write
  TTY: fix atime/mtime regression

Linus Torvalds (1):
  tty: fix up atime/mtime mess, take three

Luis Henriques (1):
  UBUNTU: ubuntu: AUFS: fsnotify_open() now receives a file

Signed-off-by: Wu Fengguang (1):
  fanotify: FMODE_NONOTIFY and __O_SYNC in sparc conflict

 drivers/char/pty.c          |  3 +++
 drivers/char/tty_io.c       | 14 ++++++++++++--
 fs/compat.c                 |  5 ++---
 fs/exec.c                   |  4 ++--
 fs/nfsd/vfs.c               |  4 ++--
 fs/open.c                   |  2 +-
 fs/read_write.c             |  8 ++++----
 include/asm-generic/fcntl.h |  9 +++++++++
 include/linux/fs.h          |  3 +++
 include/linux/fsnotify.h    | 41 ++++++++++++++++++++++++++---------------
 ubuntu/aufs/file.c          |  2 +-
 11 files changed, 65 insertions(+), 30 deletions(-)

-- 
1.8.3.2





More information about the kernel-team mailing list